When you need to union multiple query builders dynamically, array_reduce provides a clean alternative to chaining .union() calls manually. This is especially useful when the number of queries varies or comes from configuration.

The Problem with Manual Chaining

When building complex queries that combine multiple query builders with UNION, manual chaining becomes verbose:

$query1 = Order::where('status', 'pending')->toBase();
$query2 = Order::where('status', 'processing')->toBase();
$query3 = Order::where('status', 'completed')->toBase();

$combined = $query1->union($query2)->union($query3);

This gets unwieldy when:
– The number of queries changes
– Queries come from configuration
– You’re building queries conditionally

Use array_reduce Instead

array_reduce lets you build the union dynamically:

$queries = [
    Order::where('status', 'pending'),
    Order::where('status', 'processing'),
    Order::where('status', 'completed'),
];

$combined = array_reduce(
    $queries,
    fn($sub, $query) => $sub ? $sub->union($query->toBase()) : $query->toBase()
);

The closure handles the first iteration (when $sub is null) and subsequent iterations (when $sub contains the accumulated union).

Combine with fromSub for Complex Queries

This pattern shines when used with Eloquent’s fromSub():

$data = Task::query()
    ->fromSub(
        array_reduce(
            [
                Task::where('priority', 'high'),
                Task::where('priority', 'urgent'),
                Task::where('status', 'overdue'),
            ],
            fn($sub, $query) => $sub 
                ? $sub->union($query->toBase()) 
                : $query->toBase()
        ),
        'tasks'
    )
    ->with('project', 'assignee')
    ->get();

This gives you a clean subquery with proper eager loading.

Works with Any Number of Queries

The real power is flexibility:

// Configuration-driven queries
$statusQueries = config('report.statuses')
    ->map(fn($status) => Report::where('status', $status));

$results = Report::withTrashed()
    ->fromSub(
        array_reduce(
            $statusQueries->all(),
            fn($sub, $q) => $sub ? $sub->union($q->toBase()) : $q->toBase()
        ),
        'reports'
    )
    ->orderByDesc('created_at')
    ->get();

// Conditional queries
$queries = collect([
    $request->filled('active') ? Item::where('is_active', true) : null,
    $request->filled('pending') ? Item::where('status', 'pending') : null,
    $request->filled('archived') ? Item::onlyTrashed() : null,
])->filter();

$items = Item::fromSub(
    array_reduce(
        $queries->all(),
        fn($sub, $q) => $sub ? $sub->union($q->toBase()) : $q->toBase()
    ),
    'items'
)->paginate();

The pattern keeps your code DRY when query sources change or grow.

Repository vs Service: What Goes Where in Laravel

When refactoring Laravel applications, one common anti-pattern I see is services that directly perform database queries. This violates separation of concerns and makes code harder to test and maintain.

The Problem: Bloated Service Classes

Consider a ReportService that handles business logic for generating reports. Over time, it accumulates methods like:

class ReportService
{
    public function generateReport($projectId, $startDate, $endDate)
    {
        // Business logic here
        $data = $this->fetchReportData($projectId, $startDate, $endDate);
        // More business logic
    }

    private function fetchReportData($projectId, $startDate, $endDate)
    {
        // Direct database query
        return DB::table('reports')
            ->where('project_id', $projectId)
            ->whereBetween('created_at', [$startDate, $endDate])
            ->get();
    }

    private function getProjectSettings($projectId)
    {
        // Another direct database query
        return DB::table('project_settings')
            ->where('project_id', $projectId)
            ->first();
    }
}

This service is doing two jobs: orchestrating business logic AND querying the database.

The Solution: Move Database Queries to Repositories

Repositories should handle all database access. Services should orchestrate business logic by calling repositories.

Create a Repository:

class ReportRepository
{
    public function findReportData($projectId, $startDate, $endDate)
    {
        return DB::table('reports')
            ->where('project_id', $projectId)
            ->whereBetween('created_at', [$startDate, $endDate])
            ->get();
    }

    public function getProjectSettings($projectId)
    {
        return DB::table('project_settings')
            ->where('project_id', $projectId)
            ->first();
    }
}

Clean up the Service:

class ReportService
{
    public function __construct(
        private ReportRepository $reportRepo
    ) {}

    public function generateReport($projectId, $startDate, $endDate)
    {
        // Pure business logic - no database queries
        $data = $this->reportRepo->findReportData($projectId, $startDate, $endDate);
        $settings = $this->reportRepo->getProjectSettings($projectId);
        
        // Apply business rules, transformations, etc.
        return $this->processReportData($data, $settings);
    }
}

The Rule

Services orchestrate. Repositories query.

If your service has DB::table() or Eloquent queries, move them to a repository. Your service should read like a business workflow, not a database script.

Bonus: Testing Becomes Easier

With this separation, you can mock the repository in tests:

public function test_generates_report()
{
    $mockRepo = Mockery::mock(ReportRepository::class);
    $mockRepo->shouldReceive('findReportData')->once()->andReturn(collect([...]));
    
    $service = new ReportService($mockRepo);
    $result = $service->generateReport(1, '2024-01-01', '2024-12-31');
    
    $this->assertInstanceOf(Report::class, $result);
}

Clean separation = easier testing + clearer code architecture.

When building systems with both user interfaces and internal logic, explicitly separating display names from system identifiers prevents cascading changes when marketing decides to rename a feature.

The pattern: use stable identifiers (type, id, slug) for code/database/filenames, and store display names (name, label) separately in configuration:

// config/features.php
return [
    'available_reports' => [
        [
            'type' => 'sales_report',               // Stable system identifier
            'name' => 'Quarterly Sales Analysis',   // User-facing display name
            'permission' => 'view_sales_reports',
        ],
        [
            'type' => 'inventory_report',
            'name' => 'Stock Level Summary',
            'permission' => 'view_inventory',
        ],
    ],
];

In your controllers:

class ReportController
{
    public function index()
    {
        $reports = collect(config('features.available_reports'))
            ->filter(fn ($report) => auth()->user()->can($report['permission'] ?? ''))
            ->map(fn ($report) => [
                'id' => $report['type'],      // Internal ID for API/routes
                'label' => $report['name'],   // Display name for UI
            ]);
            
        return view('reports.index', compact('reports'));
    }
    
    public function generate(string $reportType)
    {
        // Use 'type' for routing, job dispatch, filename generation
        $job = match ($reportType) {
            'sales_report' => new GenerateSalesReport(),
            'inventory_report' => new GenerateInventoryReport(),
            default => throw new InvalidArgumentException(),
        };
        
        dispatch($job);
    }
}

Job classes use the stable identifier:

class GenerateSalesReport implements ShouldQueue
{
    public function handle()
    {
        $filename = 'sales_report_' . now()->format('Y-m-d') . '.pdf';
        
        Storage::put("exports/{$filename}", $this->generatePdf());
        
        // Filename: 'sales_report_2024-03-05.pdf' — never changes
    }
}

Why this matters:

• Marketing freedom: “Quarterly Sales Analysis” can become “Revenue Insights Dashboard” without touching code
• Stability: Database queries, API endpoints, and filenames don’t break when display names change
• A/B testing: Easily test different labels for the same feature
• Internationalization: Display names can be translated while system identifiers stay English

What NOT to do:

// ❌ DON'T couple display names to class constants
class GenerateSalesReport
{
    public const DISPLAY_NAME = 'Quarterly Sales Analysis';
    
    public function getFilename()
    {
        return self::DISPLAY_NAME . '_' . now()->format('Y-m-d') . '.pdf';
        // Filename: 'Quarterly Sales Analysis_2024-03-05.pdf' — spaces, changes when label changes
    }
}

// ❌ DON'T hardcode display names in multiple places
// Controller
$reportName = 'Quarterly Sales Analysis';

// Blade view
Quarterly Sales Analysis

// Email notification
Mail::send(..., ['report' => 'Quarterly Sales Analysis']);

// Now you have 3+ places to update when marketing changes the name

The right approach:

• Store display names in config/*.php or database tables where non-developers can update them
• Use system identifiers everywhere in code (sales_report, not "Quarterly Sales Analysis")
• Fetch display names at runtime from the centralized source

Your codebase becomes more flexible, and non-technical stakeholders can update user-facing labels without opening a pull request.

When generating filenames from dynamic input—user-provided names, database values, or API responses—a three-step Laravel string helper chain ensures clean, filesystem-safe output.

The pattern combines Str::lower(), Str::slug(), and Str::title() to handle edge cases you might not think of:

use Illuminate\Support\Str;

class FileGenerator
{
    public static function sanitizeFilename(string $name): string
    {
        return Str::title(Str::slug(Str::lower($name), '_'));
    }
}

// Examples
FileGenerator::sanitizeFilename('Q4 Sales Report');
// Returns: 'Q4_Sales_Report'

FileGenerator::sanitizeFilename('User Analytics (2024)');
// Returns: 'User_Analytics_2024'

FileGenerator::sanitizeFilename('Employee List - HR Dept.');
// Returns: 'Employee_List_Hr_Dept'

FileGenerator::sanitizeFilename('Données françaises');
// Returns: 'Donnees_Francaises'

Why this three-step chain works:

1. Str::lower() normalizes case to avoid filesystem issues on case-sensitive systems (Linux servers are case-sensitive, Windows/Mac are not)
2. Str::slug() converts to URL-safe format, replacing spaces and special characters with your chosen separator (underscore here)
3. Str::title() capitalizes words for readable filenames without breaking filesystem compatibility

Real-world usage with timestamps:

class ReportExporter
{
    public function export(string $reportName, array $data): string
    {
        $filename = FileGenerator::sanitizeFilename($reportName) 
                    . '_' . now()->format('Y-m-d_His') 
                    . '.csv';
        
        Storage::put("exports/{$filename}", $this->toCsv($data));
        
        return $filename;
    }
}

// Output: 'Monthly_Revenue_2024-03-05_093045.csv'

Why not just use Str::slug() alone?

Plain Str::slug() would give you 'q4-sales-report' (all lowercase). The Str::title() step makes filenames more readable when users download them. Compare:

• Without title case: employee_performance_report.csv
• With title case: Employee_Performance_Report.csv

The second is clearer at a glance in file explorers.

Alternative separators:

You can use hyphens instead of underscores by changing the second parameter to Str::slug():

return Str::title(Str::slug(Str::lower($name), '-'));

This is common for web-facing URLs. Underscores are traditional for downloaded files, but both work fine for filesystems.

If you’re building multiple Laravel AI agents, you’ll notice a lot of repetitive setup code in every agent class:

// ❌ Every agent repeats the same middleware setup
class CustomerSupport implements Agent, HasMiddleware
{
    use Promptable;

    public function middleware(): array
    {
        return [
            RateLimitMiddleware::class,
            LoggingMiddleware::class,
            RetryMiddleware::class,
        ];
    }
}

class ReportGenerator implements Agent, HasMiddleware
{
    use Promptable;

    public function middleware(): array
    {
        return [
            RateLimitMiddleware::class,
            LoggingMiddleware::class,
            RetryMiddleware::class,
        ];
    }
}

DRY it up with a base agent:

// app/Ai/Agents/BaseAgent.php
abstract class BaseAgent implements Agent, HasMiddleware
{
    use Promptable;

    public function middleware(): array
    {
        return [
            RateLimitMiddleware::class,
            LoggingMiddleware::class,
            RetryMiddleware::class,
        ];
    }
}

Now every agent just extends BaseAgent:

// ✅ Clean and minimal
class CustomerSupport extends BaseAgent
{
    // Just your agent-specific logic
}

class ReportGenerator extends BaseAgent
{
    // Just your agent-specific logic
}

Benefits:

• Add/remove middleware in one place – affects all agents instantly
• Onboarding a new agent? One line: extends BaseAgent
• Override middleware for specific agents if needed (just override the middleware() method)

Example override for special cases:

// Most agents use BaseAgent defaults
class BulkProcessor extends BaseAgent
{
    // Different rate limit for bulk operations
    public function middleware(): array
    {
        return [
            BulkRateLimitMiddleware::class,
            ...parent::middleware(),
        ];
    }
}

Pattern applies beyond agents: Same concept works for controllers (BaseController), jobs (BaseJob), or any class hierarchy where you have shared setup.